Low-code and no-code are having their moment. We build what those platforms cannot carry: whole departments replaced by systems that run millions of operations a day across dozens of internal systems — including the ones with no API and no documentation.
Twelve years, 400+ automated processes, 12M+ operations executed a day. Have a task or a problem you need solved? Send it over, whatever its size — a great many of our clients arrived with something small and stayed for years. You get an answer within 24 hours.
processes automated end to end
systems integrated, many without APIs
operations a day on a single system
years on processes others refused
We automate the flows that cross five departments and seven systems, carry real money or real liability, and have spent years being held together by people, spreadsheets and institutional memory.
Intake, validation, pricing rules, stock allocation, invoicing and dunning across systems that were never designed to talk to each other.
Volume · exceptions
Contracts, invoices, customs and shipping paperwork arriving as scans, mail bodies and thirty inconsistent supplier formats.
Unstructured input
The department whose entire job is retyping data from one screen into another because the two systems have no interface.
Legacy · no API
Classification, routing, evidence gathering and decisioning at volumes where a queue is measured in weeks, not hours.
Judgement at scale
Supplier catalogues, quotes, three-way matching and approval chains that stall on one person's inbox for four days.
Approvals · matching
Month-end closes assembled by hand from exports, where the reconciling itself is the job and the errors surface a quarter later.
Control totals · audit
Regulated multi-step workflows with evidence requirements, expiry clocks, four-eyes approval and a full audit trail.
Regulated · traceable
The one a low-code team returned as impossible. Two weeks, fixed fee, a working pilot on your real systems.
Bring it to us
We work inside systems the client owns or is authorised to operate, every action is logged and reversible, and nothing goes live without a rollback path and a measured shadow run against the humans doing it today — the same rules on every engagement, written into the contract.
Zapier, Make, Power Automate and n8n are good tools, and we recommend them when they fit. They fail on the same six things every time — and those six things are exactly what a department-scale process is made of. Here is all six, and what we do about each.
Per-operation pricing, execution timeouts and concurrency caps make a platform that is cheap at 10 000 runs a month absurd at ten million a day — and it stops long before the invoice does.
What we doEngineered services with the throughput and cost per operation as design inputs, running on infrastructure sized for the load rather than on someone else’s per-task meter.
Real processes wait days for an approval, resume after a failure and must never post an invoice twice. A chain of stateless triggers has no way to guarantee any of that.
What we doExplicit state machines with idempotent steps, compensating transactions and replay — a step can fail at any point without leaving the business half-done.
The demo automates the 80% that is easy. The 20% left over is where the cost, the risk and the entire headcount actually live — and it quietly routes back to a human inbox.
What we doException taxonomy built from your real history, with an operator console, confidence thresholds and a measured straight-through rate that we commit to.
A 1998 ERP, a terminal application, a licence-locked accounting system, a supplier portal with no interface. No marketplace connector is coming for any of them.
What we doWe build the adapter: internal APIs, direct database work, protocol-level integration, and controlled UI automation only where nothing else exists.
“Nina knows which of these to reject.” The logic is real, it is worth money, and it exists only as fifteen years of one person's pattern-matching over messy input.
What we doWe extract the decision from history and from the people who make it, encode it as versioned policy, use models only where they earn their place, and keep every decision explainable.
Business-critical logic spread across 300 opaque scenarios in a vendor's cloud, editable by anyone, with no version control, no test suite and no answer to “why did it decide that in March?”
What we doVersioned code, infrastructure as code, tests, versioned policies and a per-decision audit trail. Every decision is reviewable, and we keep it that way for as long as we run it.
Automating one task is a demo. The engineering is in carrying an entire function — including its worst days, its month-end peak and its exceptions — for years, at a cost per operation that makes the change worth making.
Operating envelope
Not a helper beside the team — the process itself, running unattended, with humans left only on the decisions that genuinely need them. Volume, peak behaviour and the exception budget are fixed on day one, not discovered in month four when the queue starts growing.
Where the volume goes
Illustrative profile of one production system. Your envelope is fixed in the design phase and written into the SLA.
Every automation we ship runs on the same contour. Blocks change per process; the property that does not change is that a failure anywhere is detected, contained and replayed — never silently posted into a production system.
The difference between an integrator and an automation engineering team shows up months after go-live — in whether the process still runs unattended, and whether anyone can prove what it did and why.
Every process looks simple until you count the ways it goes wrong. We start from the last twelve months of real cases — the malformed input, the missing counterparty, the supplier who sends a photograph of a fax — and design for that distribution rather than the version described in the process document. That is why our straight-through rates hold in month nine.
The value is not in shaving twenty minutes off a job. It is in a function that runs unattended: intake, decision, execution and control, with people left on the judgement calls that actually deserve a person. We agree the target operating model with you before the build, including what happens to the team — reassignment, attrition or reduction is your decision, made openly rather than discovered.
At department scale, the recurring line — compute, document processing, model calls, licences — decides whether the automation is worth having. We design the execution model per process, use the expensive machinery only where it earns its place, and report cost per thousand operations from day one.
Tests, runbooks, infrastructure as code, versioned policies and a documented decision model. Shipping is the start of it, not the end: hosting, monitoring, on-call, upstream changes, tuning and new features all sit inside one monthly arrangement, handled by the same engineers who designed the thing. A system we will still be running in year eight has to be engineered to be operated, not merely delivered — and most of our clients are on their third or fourth system with us.
An automation that carries a department is a production system with an uptime, not a project that was delivered once. These are the commitments in the support agreement; tiers and windows are set per project against how critical the process is.
| Commitment | Target | What it means in practice |
|---|---|---|
Straight-through rate |
95 %+ | The share of cases completed with no human touch, measured continuously against the agreed target rather than asserted. When it drops, you see it on the same dashboard we do. |
Process monitoring |
24 / 7 | Queue growth, SLA breaches, error spikes, decision drift and silent stalls are detected by us and raised to you — not discovered by an operations manager at month-end. |
Action integrity |
Every run | Idempotent execution, reconciliation against the target systems, and a full per-step audit trail. A failing case is contained and flagged, never half-posted. |
Engineering response |
From 4 h | Incident tiers with named windows. A system upgrade, a changed form, a new supplier format or a policy change is routine work covered by the retainer, not a change request. |
Recovery |
Defined RTO | Checkpoints, compensation paths, replay queues and a documented manual fallback, so an upstream outage costs a window of latency — not a week of untraceable state. |
Availability |
99.9 % | Measured on completed runs against schedule and reported monthly. You get the same dashboard we watch. |
Each of these came to us after a low-code build stalled, an RPA farm became unmaintainable, or an integrator returned the process as impossible. Clients are under NDA — the engineering detail we walk through against your own case.
Shipping, customs and supplier paperwork in 40+ formats, parsed, checked against the ERP and posted, for a European freight group.
What was hardScans and photographs rather than data, per-country document rules, a customs system with no interface, and a rejection cost measured in detained containers.
Full commercial cycle — intake, credit checks, allocation, invoicing and dunning — unified across an ERP, two CRMs, a WMS and a 20-year-old accounting core.
What was hardNo API on the accounting core, contradictory master data between systems, month-end peaks at eight times average load, and a previous low-code build that had been double-posting invoices for months.
Intake, classification, evidence gathering and first-line decisioning for a claims department of 60, with full audit traceability for the regulator.
What was hardDecision logic that existed only in the heads of four senior assessors, free-text and photographic evidence, and a regulatory requirement to explain every automated decision.
On a real department process, a quote written from a brief is a guess — the spread between the described version and the actual one is routinely tenfold. So we prove it on your real systems first, for a fixed fee.
1Brief
Which process, which systems, what volume, what the exceptions actually look like, and what changes for the business if it runs itself. Half the time the scope narrows and gets cheaper right here.
OutputProcess map & exception profile
2Pilot
Two to four weeks on your real systems and your real cases: one slice running end to end, measured straight-through rate, measured cost per operation — and the honest risks.
OutputRunning pilot + risk list
3Design
Target operating model, exception handling, decision policy, throughput, straight-through target and running cost — written into the contract, not into a slide.
OutputFixed scope & fixed price
4Build
Every week ships something that runs in your environment, shadowing the current team before it takes over. Monitoring, runbooks and documentation are built alongside it, not bolted on at the end.
OutputA system running in production
5Operate
On-call, handling system upgrades and changed formats, policy tuning and extensions. Routine changes sit in the retainer; large ones are quoted first.
OutputA process you stop thinking about
Automation prices by exception complexity and by how hostile the systems are, not by step count. So we price in two steps, and we say no early when the work is not ours.
Model
You start on whichever rung fits, and most clients start with one small first task. From the Operate rung onward it is a single monthly arrangement covering hosting, monitoring, on-call and continued development, so nobody on your side has to build an operations team around it. Infrastructure runs on ours or inside your own perimeter — whichever your security and finance people prefer.
Not our work
If your process is real but not our kind of work, we will say so in the first reply and point you at someone better suited — including a low-code platform when that is genuinely the right answer. That costs us nothing and saves you a quarter.
Legal & governance
Systems act only inside the client's own perimeter and under credentials the client authorises, with segregation of duties preserved, every action logged and every automated decision explainable. Access, retention and deletion controls are designed into the process rather than bolted on.
GDPR alignment and internal-control requirements are implemented through technical and organisational controls. Confirming the lawful basis, the regulatory treatment and the approval authority for a specific process and jurisdiction remains with the client’s legal and risk teams — we give them the documentation to do it.
If yours isn’t here, put it in the form — we answer in writing, within 24 hours, and without a discovery call first. Most questions get a straight yes or no rather than a proposal.
Bring us the hard process
Send the process, the systems it touches and the volume it carries. You get a straight answer: whether it is automatable, what specifically makes it hard, and roughly what it costs to build and to run. No deck, no discovery call before there is anything to discover.